A case-control study on occupational risk factors for sino-nasal cancer

Transcription

1 1 Epidemiology Unit, Piedmont Region, Grugliaso, Italy; 2 Oupational Mediine Department, Savigliano, Italy; 3 Oupational Health and Safety Department, Asti, Italy; 4 Oupational Health and Safety Department, Saluzzo, Italy; 5 Doumentation Center for Health Promotion, Grugliaso, Italy; 6 Otolaryngology (ENT) Unit, Giovanni Boso Hospital, Turin, Italy; 7 University of Turin, Department of Publi Health, Turin, Italy Correspondene to: A d Errio, Loal Health Unit ASL TO3, Epidemiology Department - Piedmont Region, Via Sabaudia 164, Grugliaso (TO) 10095, Italy; angelo.derrio@ epi.piemonte.it Aepted 13 November 2008 Published Online First 19 January 2009 This paper is freely available online under the BMJ Journals unloked sheme, see oem.bmj.om/info/unloked.dtl A ase-ontrol study on oupational risk fators for sino-nasal aner A d Errio, 1 S Pasian, 1 A Baratti, 2 R Zanelli, 3 S Alfonzo, 4 L Gilardi, 5 F Beatrie, 6 A Bena, 1 G Costa 7 ABSTRACT Objetives: Sino-nasal aner has been onsistently assoiated with exposure to wood dust, leather dust, nikel and hromium ompounds; for other oupational hazards, the findings are somewhat mixed. The aim of this study was to investigate the risk of sino-nasal epithelial aner (SNEC) by histologial type with prior exposure to suspeted oupational risk fators and, in partiular, those in metalworking. Methods: Between 1996 and 2000, inident ases were olleted on a monthly basis from hospitals throughout the Piedmont region of Italy by the regional Sino-nasal Caner Registry. A questionnaire on oupational history, ompleted by 113 ases and 336 hospital ontrols, was used to assign exposure to oupational hazards. The relationship between SNEC and umulative exposure to these hazards was explored using unonditional logisti regression to statistially adjust for age, sex, smoking and o-exposures, allowing for a 10-year lateny period. Results: The risk of adenoarinoma was signifiantly inreased with ever-exposure to wood dust (odds ratio; OR = 58.6), and to leather dust (OR = 32.8) and organi solvents (OR = 4.3) after ontrolling for wood dust, whereas ever-exposure to welding fumes (OR = 3.7) and arseni (OR = 4.4) signifiantly inreased the risk for squamous ell arinoma. For eah of these hazards, a signifiant inreasing trend in risk aross ordered umulative exposure ategories was found and, exept for arseni, a signifiantly inreased risk with everexposure at low intensity. Treating umulative exposure on a ontinuous sale, a signifiant effet of textile dusts was also observed for adenoarinoma. For a mixed group of other histologial types, a signifiant assoiation was found with wood dust and organi solvents. Conlusions: Some oupational risk fators for SNEC were onfirmed, and dose response relationships were observed for other hazards that merit further investigation. The high risk for adenoarinoma with low-intensity exposure to wood dust lends support for a redution in the oupational threshold value. Malignant tumours of the nose and paranasal sinuses (SNC) are unommon neoplasms with an inidene of less than one per inhabitants in most developed ountries. 1 Over three-quarters of all SNC onsist of malignant epithelial tumours, among whih the most ommon histologies are adenoarinoma and squamous ell arinoma. 2 A distint harateristi of SNC and, in partiular, of epithelial tumours is that the major risk fators are workplae hazards. Sine the late 1960s, many epidemiologial studies have found a high risk of SNC among employees in woodworking and furniture prodution, 3 5 nikel refining 6 7 and the manufature of shoes and other leather produts. 8 9 The assoiation between adenoarinoma and wood dust exposure, espeially hardwood, is well established and ontributed to its lassifiation by the International Ageny for Researh on Caner (IARC) as definitely arinogeni to humans. 10 For other histotypes, the relationship appears to be less onsistent and the risk is muh lower. 11 Nikel ompounds are aknowledged risk fators for SNC, and are also lassified as definitely arinogeni to humans. 12 The risks for SNC observed in shoe and leather manufaturing are likely attributable to leather dust exposure, as the risk is espeially high among the most heavily exposed workers employed in finishing 8 and there is evidene of a dose response effet. 9 Nonetheless, only the boot and shoe manufature and repair setor has been lassified as a group 1 arinogen and not leather dust itself, as the hemials used in leather manufaturing may have been responsible for the exess risks. The assoiation with leather dust appears mainly with adenoarinoma, although an inreased risk has been reported for other histologies Oupational exposure to hexavalent hromium ompounds, also definite human arinogens, has been onsistently linked to SNC. 12 Inreased risks of SNC have also been found, with some inonsistenies, for many other oupations, inluding farm workers, 15 textile and hemial 18 workers, bakers and pastry onfetioners, 19 onstrution workers, 15 welders 20 and other metal workers. 21 The rarity of SNC makes it unlikely that exess risks will be observed in ohort studies; exept for a few ohorts onsisting mainly of workers employed in the wood and leather industries or in nikel refining, most of the evidene relating SNC to oupational exposures omes from aseontrol studies in whih the risk estimates are often based on small numbers of exposed ases, and the prevalene of exposure to most suspeted hazards is low. The aim of this study was to investigate the risk of sino-nasal epithelial aner (SNEC) with prior exposure to suspeted oupational risk fators, in the Piedmont region of Italy. As metalworking was one of the most ommon oupations in Piedmont for many deades after the Seond World War, the prevalene of exposure to the known or suspeted arinogeni agents to whih workers would have been exposed in this industry was suffiiently high to examine the risks by histologial types. 448 Oup Environ Med 2009;66: doi: /oem

2 METHODS Starting in 1996, all Piedmont hospital departments in whih SNC ould be diagnosed or treated were ontated monthly by the Piedmont SNC Registry, established for the identifiation of oupational-related SNC ases. By 2000, 154 inident ases of primary SNEC (International Statistial Classifiation of Diseases and Related Health Problems, 9th revision: 160.0, ) with histologial onfirmation had been identified in 22 otolaryngology (ear, nose and throat; ENT) and two maxillofaial surgery departments. A rude inidene rate of 0.73 SNEC ases per inhabitants was estimated using the 2001 Census data for the denominator (1.19 per among men and 0.30 per among women). Forty-one ases were exluded: three prevalent ases; 36 not interviewed or interviewed using a pilot version of the questionnaire; and two under 30 years old, in onsideration of the long lateny of SNC tumours. 22 Among the hospital ontrols ontated between 1998 and 2002, the non-response rate was 5%, and six were exluded (four not interviewed and two under 30 years old). The 336 ontrols were reruited from departments of ENT and orthopaedis, frequeny mathed to ases by age (10-year lasses), sex and provine of residene. Study partiipants ompleted a 1 h open-ended questionnaire administered by trained interviewers (not blind to ase-ontrol status) on job history, demographis and tobao smoking. For eah working period, defined as a ontinuous period of employment in the same job, partiipants were asked to report on: ompany size and type of eonomi ativity; job title; desription, frequeny and duration of their tasks; size of their work environment, number of people working in it and the tasks performed by them; desription of mahinery in their work environment; substanes used by themselves and others in their proximity; proportion of working hours spent outdoors; presene of dusts, fumes and vapours in their work environment; use of protetive devies; and the presene and effiieny of exhaust ventilation systems. All subjets gave their informed onsent to partiipate in this study. The list of possible risk fators for SNC inluded arseni, wood dust, leather dust, nikel ompounds, hromium VI and its salts, polyyli aromati hydroarbons (PAHs), welding fumes, oil mists, formaldehyde, flour, ooa powder, textile dusts, silia, oal dust, paint mists, strong aid mists and organi solvent vapours. Wood dust was not separated into hardwood or softwood as this information was olleted only in terms of lifetime oupational exposure. For eah working period of at least 6 months, the probability and intensity of exposure to eah hazard were independently rated based on the information provided in the questionnaires, by an oupational epidemiologist and by one of two oupational physiians; in the event of non-onordant ratings (38%), the questionnaire was re-evaluated by a senior industrial hygienist, all of whom were blind to ase-ontrol status. In addition, job title (ISCO-68) and eonomi setor (ISIC-74) odes, assigned by an expert, were used in onjuntion with two job-exposure matries (JEMs) to aid in evaluating some exposures: solvent vapours, arseni, PAHs, hromium and nikel ompounds ; wood, leather, textile, oal and silia dusts, flour, formaldehyde. 25 For these hazards, the probability and intensity provided in the JEM by eonomi setor and job title were adjusted in aordane with information in the questionnaire (historial period and working onditions), expert opinion of the raters and, in some instanes, industrial hygiene reports. The probability of exposure was rated on a four-level sale, in whih unexposed orresponds roughly to less than Original artile 10%, low to between 10 and 50%, medium in the range 50 90% and high to a probability above 90%. The intensity rating was either negligible (unexposed), low, medium or high. A working period was defined by its start and end year in whih mid-year was used to alulate the duration of exposure in years. A lateny period of 10 years was assumed and, therefore, any exposure ourring within 10 years of diagnosis was exluded. For eah hazard, the duration of exposure was weighted by the probability and intensity of exposure, and umulative exposure was obtained by summing the weighted duration over all working periods. As shown in table 1, highintensity exposures had all been rated as high probability, and medium intensity as either medium or high probability. In seleting the weights, intensity was onsidered as two levels (low and medium high) given the sparseness of data. Exposures that had been rated as high probability and either medium or high intensity were weighted at 100%, suh that every year of exposure ounted toward umulative exposure. The weight assigned to exposures rated as low probability, all of whih were at low intensity, was 30%, the mid-point of the probability range. One year of umulative exposure, for instane, orresponds to 3.5 years of low-intensity exposure if at low probability, 2 years if at medium and 1.5 years if rated as high probability. Statistial analysis The relationship between the probability of SNEC and umulative exposure to oupational hazards was explored using unonditional logisti models to estimate relative risks statistially adjusted for o-exposures; this relationship was examined further by histologial type (adenoarinoma, squamous ell arinoma and a mixed group of other histotypes). In a preliminary analysis, odds ratios (ORs) for ever-exposure to hazards were adjusted by age and sex, as there was no interation with sex. The ever-exposed ategorisation was defined as umulative exposure of at least 1 year. Given the small number of exposed ases and the inherent loss of information in ategorising quantitative variables, multivariable models were fitted treating umulative exposure to all hazards as ontinuous variables, with the assumption that the multipliative hange in risk for every unit inrease in years of exposure is onstant over the range of eah hazard. The relative risk estimates obtained using unonditional logisti regression to model ontinuous data from ase-ontrol studies have been shown to be relatively robust over the range to whih 90% of the population are exposed, even if the data are inorretly modelled as ontinuous variables. 26 The independent variables were sex, age (ontinuous), smoking status (never, former, urrent) and those hazards with at least three exposed ases. In two subsequent analyses, umulative exposure was defined as ategorial variables: (1) dihotomous (ever/never exposed) Table 1 Years of exposure weighted by probability and intensity equivalent to 1 year of umulative exposure* Intensity Low Medium High Duration Probability Assigned weights Years of exposure 10 50% Low % Medium % High *Assigned weight 6 years of exposure = umulative exposure. Oup Environ Med 2009;66: doi: /oem

3 Table 2 Distribution of sino-nasal epithelial aner by age, sex and smoking status Adenoarinoma Squamous ell arinoma Other histotypes Controls Age (years) No (%) No (%) No (%) No (%) (5.7) 6 (16.2) 1 (4.4) 51 (15.2) (7.5) 6 (16.2) 5 (21.7) 62 (18.5) (49.1) 15 (40.6) 9 (39.1) 109 (32.4) >70 20 (37.7) 10 (27.0) 8 (34.8) 114 (33.9) Sex Males 45 (84.9) 12 (32.4) 19 (82.6) 234 (69.6) Females 8 (15.1) 25 (67.6) 4 (17.4) 102 (30.4) Smoking Never 18 (34.0) 11 (29.7) 10 (43.5) 128 (38.0) Ex-smoker 18 (34.0) 9 (24.3) 9 (39.1) 102 (30.4) Current 16 (30.1) 15 (40.6) 4 (17.4) 102 (30.4) Missing 1 (1.9) 2 (5.4) 0 (0.0) 4 (1.2) Total 53 (100.0) 37 (100.0) 23 (100.0) 336 (100.0) and (2) polyhotomous (three ordered exposure ategories unexposed and two lasses hosen aording to the range of eah hazard in the sample data). Finally, the effet of exposure intensity was evaluated separately, weighting duration of exposure only by probability (0.3, 0.7 and 1.0 for low, medium and high, respetively) and treating intensity as a ategorial variable (unexposed, exposed only at low intensity, ever exposed at medium high intensity). The same strategy was followed to selet a best subset model in eah analysis, inluding an assessment of multiollinearity. Mallows Cp statisti, whih is a measure of predition error, was used to hoose among all possible regression models with preferene given to those with the smallest Cp values. As these models tend to ontain the most important preditors and minimise the error in predition, they were assessed in seleting the final models. 27 First-order interations were assessed in the models with ategorial exposure variables. All tests were twosided, for whih a signifiane level of 0.05 was adopted. SAS (version 8) and STATA (version 8) were used to perform the analyses. The effets of potential onfounding variables were onsidered; a onfounder was retained in the model if its removal resulted in a hange of 15% or more in any regression oeffiient. In the models with ategorial variables, onfounding by wood dust was assessed by adjusting for it on a ontinuous sale. The presene of a trend in risk aross ordered exposure lasses was examined by treating these variables as ontinuous. Diagnosti heks on model adequay inluded verifying that the logit was linear in the ontinuous independent variables. The stability of the final regression models was asertained by restriting the analyses to exposures that had ourred with a probability of at least 50% (medium and high). The final models for adenoarinoma were also ompared with those obtained when the analysis was restrited to those persons who had not been exposed to leather dust, whih exluded four ases and two ontrols. It was not possible to do the same for wood dust, beause the remaining ases were too few in number to examine the effets of other hazards. RESULTS Of the 113 ases, 53 were adenoarinomas (ADs), 37 squamous ell arinomas (SCCs) and 23 were other histologies (OHs), inluding 11 undifferentiated, six muoepidermoid, one neuroendorin, one basoellular and four unspeified arinomas (table 2). Cases and ontrols did not differ signifiantly in the average number of years worked in ISCO 1-digit oupational groups 5 to 9, whih onsist mainly of manual workers (32.9 vs 30.1 years, t test: p = 0.14). The ourrene of any SNEC was signifiantly related to everexposure to wood dust (OR = 11.4) in the age sex-adjusted analyses (table 3). The risk for AD with wood dust exposure (OR = 58.6) was more than 10-fold that observed for OH, and these risk estimates were virtually unhanged after ontrolling for ever-exposure to other signifiant o-exposures (not shown). In table 4, ORs are presented for the final multivariable models, in whih umulative exposure was treated first as ontinuous and then as ategorial variables (ordered exposure lasses and ever exposed). The risk for AD more than doubled with every 5- year inrease in exposure to wood dust (p,0.0001); for example, with 20 years of exposure the risk was 25 times higher than for those unexposed to wood dust after adjusting for o-exposures. A similar inreasing effet was found on the risk of OH (p = ), but the dose response relationship was less pronouned. A signifiant inrease in the risk of AD (OR = 16.6) was also observed with ever-exposure at only low intensity (table 5). The age sex-adjusted risk for any SNEC (OR = 14.4) with ever-exposure to leather dust was unhanged (OR = 14.3) after adjusting for wood dust and ever-exposure to other hazards (table 4), whereas the risk of AD (OR = 26.6) inreased slightly (OR = 32.8) after ontrolling for wood dust and solvent vapours. A signifiant dose response relationship was found between the risk of AD and years of leather dust exposure on a ontinuous sale (p = 0.003) and aross ordered exposure lasses (p,0.0002), with an almost 60-fold inrease in risk with over 5 years exposure relative to unexposed. As with wood dust, low-intensity exposure signifiantly inreased the risk for AD (OR = 52.4). The signifiant twofold inrease in the age sex-adjusted risk for any SNEC with exposure to welding fumes was limited to SCC (OR = 4.1) and was essentially unhanged (OR = 3.7) after aounting for arseni (OR = 4.4) (table 4). The risk of SCC inreased by about 50% with every 5-year inrease in exposure to welding fumes (p = 0.007). The trend in risk aross ordered exposure lasses was also signifiant (p = 0.001), as was the risk (OR = 3.5) with low-intensity exposure (table 5). The age sex-adjusted risk for any SNEC with ever-exposure to organi solvents (OR = 4.3) dereased but remained signifiant after ontrolling for exposure to wood dust and other hazards (OR = 2.2). A similar derease was also observed in the 450 Oup Environ Med 2009;66: doi: /oem

4 Table 3 Odds ratios (ORs) for any sino-nasal epithelial aner (SNEC) and three histologial types with ever-exposure to oupational hazards, adjusted for age and sex All SNEC Adenoarinoma Squamous ell arinoma Other histotypes Cases (n = 113) Controls (n = 336) Adjusted odds ratios Cases (n = 53) Adjusted odds ratios Cases (n = 37) Adjusted odds ratios Cases (n = 23) Adjusted odds ratios Hazard No No OR (95% CI) No OR (95% CI) No OR (95% CI) No OR (95% CI) Arseni (0.69 to 6.69) (0.08 to 5.42) (1.20 to 22.20)* (0.59 to 17.26) Wood dust (6.29 to 20.74)*** (23.74 to 144.8)*** (0.19 to 3.83) (1.99 to 15.24)*** Leather dust (3.03 to 68.87)*** (5.09 to 139.0)*** (0.44 to 56.83) (0.54 to 73.09) Nikel Chromium (0.55 to 14.06) (0.22 to 21.1) (1.40 to 60.13)* PAHs (0.37 to 1.15) (0.15 to 0.94)* (0.54 to 2.81) (0.21 to 2.05) Welding fumes (1.00 to 3.82)* (0.52 to 3.52) (1.66 to 10.13)** (0.22 to 4.66) Oil mists (0.32 to 1.40) (0.11 to 1.33) (0.43 to 3.39) (0.14 to 2.84) Formaldehyde (1.32 to 14.10)* (2.62 to 34.20)*** (0.38 to 36.53) Flour dust (0.19 to 2.48) (0.06 to 3.68) (0.34 to 7.43) 0 Cooa powder (0.25 to 70.53) (0.56 to 164.8) 0 Silia (0.21 to 1.19) (0.19 to 1.70) (0.12 to 2.28) (0.04 to 2.52) Coal dust (0.14 to 3.32) (0.08 to 5.42) (0.17 to 13.0) Textile dusts (0.68 to 2.97) (0.70 to 5.10) (0.12 to 2.34) (0.84 to 9.42) Aid mists (0.61 to 4.90) (0.27 to 6.27) (1.75 to 22.49)** Paint mists (1.47 to 6.76)** (2.23 to 12.64)*** (0.26 to 5.46) (0.43 to 9.75) Organi solvents (2.62 to 7.20)*** (4.32 to 15.72)*** (0.39 to 2.91) (2.28 to 14.21)*** *p(0.05. **p(0.01. ***p( PAH, polyyli aromati hydroarbon. Oup Environ Med 2009;66: doi: /oem

5 Table 4 Odds ratios (ORs) for any sino-nasal epithelial aner (SNEC) and three histologial types with umulative exposure (CE) to oupational hazards, adjusted for o-exposures CE treated as ontinuous variables CE treated as ategorial variables OR (95% CI) at 5-year intervals over the range for whih there were exposed ases and ontrols Ordered exposure lasses Ever exposed All SNEC 5 years 10 years 15 years 20 years 25 years 30 years 35 years OR* (95% CI) OR* (95% CI) Wood dust Leather dust Solvent vapours Welding fumes 1.77 (1.46 to 2.14) 3.1 (2.13 to 4.56) 5.5 (3.11 to 9.74) 9.7 (4.53 to 20.80) 17.1 (6.61 to 44.42) 30.2 (9.64 to 94.86) 53.4 (14.07 to 202.6) 4.76 (1.21 to 18.73){ 22.7 (1.47 to 350.7){ 1 5 years Over 5 years 14.7 (1.37 to 156.7) 16.5 (1.75 to 154.7) 1.27 (1.03 to 1.56) 1.6 (1.07 to 2.43) 2.0 (1.10 to 3.78) 2.6 (1.14 to 5.88) 3.3 (1.17 to 9.16) 4.2 (1.21 to 14.27) 1 10 years Over 10 years 1.4 (0.61 to 3.26) 3.5 (1.50 to 8.09) 1.34 (1.03 to 1.75) 1.8 (1.06 to 3.04) 2.4 (1.09 to 5.31) 3.2 (1.13 to 9.27) 4.3 (1.16 to 16.17) 1 10 years Over 10 years 2.4 (0.92 to 6.38) 3.0 (1.13 to 8.00) Arseni Not signifiant 1 10 years Over 10 years 1.6 (0.31 to 8.82) 9.3 (1.50 to 57.33) Adenoarinoma Wood 2.24 (1.76 to 2.85) 5.0 (3.09 to 8.10) 11.2 (5.42 to 23.04) 25.0 (9.53 to 65.55) 55.9 (16.74 to 186.5) (29.4 to 530.7) (51.67 to.999) dust Leather 6.94 (1.96 to 24.61){ 48.2 (3.84 to 605.4){ 1 5 years Over 5 years dust 11.3 (0.91 to 141.3) 58.1 (5.84 to 579.3) Solvent 1.38 (1.07 to 1.78) 1.9 (1.15 to 3.16) 2.6 (1.24 to 5.61) 3.6 (1.33 to 9.97) 5.0 (1.42 to 17.72) 6.9 (1.53 to 31.48) 1 10 years Over 10 years vapours 3.7 (1.28 to 10.89) 5.7 (1.82 to 18.00) Textiles 1.44 (1.08 to 1.94) 2.1 (1.16 to 3.77) 3.0 (1.24 to 7.31) 4.3 (1.33 to 14.18) 6.3 (1.43 to 27.53) 9.1 (1.54 to 53.42) Not signifiant 14.3 (2.86 to 71.39) 2.2 (1.17 to 3.97) 2.7 (1.31 to 5.45) 3.4 (1.03 to 11.43) 32.8 (5.96 to 181.1) 4.3 (1.83 to 10.11) Squamous ell arinoma Welding 1.49 (1.12 to 1.98) 2.2 (1.25 to 3.94) 3.3 (1.40 to 7.81) 4.9 (1.56 to 15.50) 7.3 (1.75 to 30.76) 1 10 years Over 10 years fumes 2.6 (0.69 to 9.46) 5.4 (1.87 to 15.33) Arseni Not signifiant 1 10 years Over 10 years Other histotypes Wood 1.51 (1.22 to 1.87) 2.3 (1.49 to 3.48) 3.4 (1.82 to 6.48) 5.2 (2.22 to 12.09) 7.8 (2.70 to 22.54) 11.8 (3.30 to 42.03) 17.8 (4.02 to 78.37) 1.6 (0.16 to 15.55) 11.9 (1.61 to 88.07) dust{ Solvent 1.36 (1.04 to 1.79) 1.9 (1.08 to 3.19) 2.5 (1.12 to 5.69) 3.4 (1.16 to 10.15) 1 10 years Over 10 years vapours 2.6 (0.66 to 10.10) 6.2 (1.82 to 21.33) Aid mists 2.60 (1.35 to 5.01) 6.7 (1.81 to 25.11) 17.5 (2.44 to 125.8) 1 10 years Over 10 years 6.0 (1.32 to 26.98) 15.6 (1.30 to 186.8) 3.7 (1.59 to 8.81) 4.4 (1.04 to 18.16) 3.8 (1.45 to 10.09) 7.5 (2.01 to 28.20) *Adjusted for wood dust exposure on a ontinuous sale to redue residual onfounding (exept for squamous ell). {ORs are for 3.5 and 7 years, the latter being the maximum years of exposure among ontrols. {Exposure among ases ranged from 21 to 39 years. 452 Oup Environ Med 2009;66: doi: /oem

6 Table 5 Odds ratios (ORs) for sino-nasal epithelial aner (SNEC) by intensity of exposure to oupational hazards, adjusted for o-exposures Categorial intensity levels* (only low or ever medium high) Level OR (95% CI) All SNEC Wood dust Low 3.2 (1.36 to 7.44) High 30.6 (11.93 to 78.30) Leather dust Low 17.6 (1.84 to 168.5) High 11.9 (1.12 to 127.1) Solvent vapours Low 2.3 (1.15 to 4.46) High 2.1 (0.65 to 6.67) Welding fumes Low 3.3 (1.47 to 7.26) High 1.6 (0.34 to 7.75) Arseni{ Ever exposed 4.1 (1.21 to 13.76) Adenoarinoma Wood dust Low 16.6 (5.10 to 54.04) High (55.37 to 584.4) Leather dust Low 52.4 (3.71 to 740.2) High 68.8 (5.60 to 844.7) Solvent vapours Low 4.5 (1.54 to 12.90) High 3.8 (0.67 to 21.58) Squamous ell arinoma Welding fumes Low 3.5 (1.31 to 9.60) High 4.3 (1.01 to 18.10) Arseni{ Ever exposed 4.3 (1.01 to 18.10) Other histotypes Wood dust Low 2.1 (0.42 to 10.50) High 16.0 (3.93 to 65.24) Solvent vapours Low 4.8 (1.72 to 13.39) High 2.9 (0.33 to 24.73) Aid mists Low 5.6 (1.25 to 25.11) High 17.1 (1.42 to 206.7) *Based on years of exposure weighted only by probability. {Could not be ategorised by intensity. age sex-adjusted risks for AD and OH after ontrolling for oexposures. The dose response relationships found for AD and OH with inreasing years of exposure to organi solvents on a ontinuous sale (p = and p = 0.026) and aross ordered exposure lasses (p = and p = 0.002) were also similar, as were the signifiantly inreased risks with low-intensity exposure (OR = 4.5 and OR = 4.8, respetively). The slightly lower risk estimates at medium high intensity were based on few exposed ases (table 5). Ever-exposure to paint mists signifiantly inreased the age sex-adjusted risk for any SNEC (OR = 3.2) and, in partiular, for AD (OR = 5.3), but its effet disappeared after aounting for exposure to organi solvents. Likewise, the relationship between any SNEC and formaldehyde exposure (OR = 4.3), also found for AD (OR = 9.5), lost signifiane after ontrolling for wood dust exposure. The effet of hromium exposure on the risk of OH was signifiant in the age sex-adjusted analysis (OR = 9.2), as was the assoiation between OH and nikel exposure (Fisher s exat test; p = ). A signifiant dose response with inreasing years of exposure to strong aid mists was observed both on a ontinuous sale (p = 0.004) and aross ordered exposure lasses (p = 0.002) after ontrolling for o-exposures; however, three of the four ases exposed to aid mists had also been exposed to nikel or hromium, neither of whih ould be ontrolled for in Original artile the multivariable model (no exposed ontrols and insuffiient number of exposed ases, respetively). The relationship between AD and exposure to textile dusts was signifiant on a ontinuous sale (p = 0.015) with an inrease in risk of over 40% with every 5-year inrease in exposure. Aside from non-signifiant inreased risks for any SNEC and SCC with exposure to ooa powder, all other hazards produed non-signifiant risk estimates of less than 2 (table 3). Age, sex and smoking were not signifiant risk fators for SNEC or any histotype, after aounting for exposure to oupational hazards. When the analyses were restrited to exposures that had ourred with at least 50% probability, the same final models were obtained for ontinuous exposure variables. Furthermore, the risk estimates for AD were very similar when those persons who had been exposed to leather dust were exluded (data not shown). DISCUSSION The ourrene of any SNEC was signifiantly related to exposure to wood dust, leather dust, organi solvents, welding fumes and arseni. The first three were signifiant preditors of AD, whereas exposure to welding fumes and arseni were risk fators for SCC. Wood dust and organi solvents were also signifiantly related to OH, but the strength of the assoiation with wood dust was one order of magnitude smaller than for AD. A signifiant effet of textile dust exposure on a ontinuous sale was also observed for AD. Among the strengths of this study was the availability of aurate histologial information, whih meant that the risks for the two most ommon histologial types, AD and SCC, ould be investigated separately. Seond, as all suspeted oupational arinogens were assessed with the exeption of asbestos, the effets of o-exposures were onsidered. Exposure to asbestos, whih has been assoiated with SNC 25 and may be orrelated with exposure to other hazards suh as welding fumes, was not inluded beause preision in reonstruting oupational exposure was onsidered unobtainable, espeially for older working periods. The assoiation between SNEC and previous exposure to wood dust was onfirmed and, speifially, with AD and OH for whih dose response effets were found. The inreased risk for AD with low-intensity wood dust exposure appears to ontradit the results from a pooled analysis of 12 ase-ontrol studies by Demers et al, 11 but this disrepany is likely explained by Demers reliane on a JEM to assign exposures, a method that usually results in substantial non-differential mislassifiation. 28 In addition to the questionnaire responses, we referred to a JEM 25 in assigning wood dust exposure, one in whih intensity levels were defined by historial period, and low intensity orresponded to less than 1 mg/m 3, medium intensity to 1 5 mg/m 3 and high intensity above 5 mg/m 3. It is unlikely that an exposure at or above 5 mg/m 3, the oupational limit for wood dust in many ountries inluding Italy, would have been rated as low intensity in the present study. Therefore, our finding suggests that the urrent exposure limit for wood dust should be onsidered for a substantial redution, as proposed by the US National Institute of Oupational Safety and Health (NIOSH) as early as The absene of an assoiation between SCC and wood dust exposure is onsistent with the findings from other studies, although modestly inreased risks with previous employment in woodworking have been reported. Oup Environ Med 2009;66: doi: /oem

7 The assoiation between AD and leather dust exposure was also onfirmed. The inreasing risk of AD with years of umulative exposure supports the existene of a dose response relationship, also found in another ase-ontrol study onduted in Italy. 9 Unlike wood dust, intensity levels for leather dust were not quantitative in the JEM 25 ; nonetheless, the inreased risk with low-intensity exposure suggests that workplae exposure to leather dust should also be minimised. Several ase-ontrol studies in Nordi ountries have reported an inreased risk of SNC in oupations involving exposure to welding fumes, but none reported results stratified by histotype. In the present study, exposure to welding fumes inreased the risk for SNEC and, in partiular, for SCC with a dose response relationship that, to our knowledge, is a new finding that supports a ausal relationship and should be investigated further. Although this assoiation an be onfounded by exposure to nikel and hromium in stainless steel welding, none of the nine SCC ases exposed to welding fumes had ever been employed as stainless steel welders. In 1990, the IARC onluded that the evidene on welding fumes was limited and lassified exposure to welding fumes as only possibly arinogeni to humans (2B). 12 Sine then, exposure to welding fumes has also been linked to lung aner in several epidemiologial studies onduted in Europe and the USA, not only among stainless steel welders but also among mild steel welders and, in some instanes, the risks inreased with duration of exposure. A meta-analysis on the risk of lung aner related to welding 33 revealed signifiantly inreased risks among both stainless and mild steel welders with similar relative risks, albeit the extent of onfounding by asbestos exposure was not asertained. Among the other signifiant findings, the inreased risks for AD and OH with exposure to organi solvents were similar in magnitude after ontrolling for other signifiant hazards, primarily wood dust. In the ase of AD, the assoiation may be attributable to residual onfounding by wood dust owing to the strength of its effet, but very similar dose response relationships were observed for both histogroups. This apparent assoiation warrants further investigation, and as organi solvents inlude many hemials, their individual effets should be examined. We are aware of only one report of an exess risk assoiated with organi solvents, 20 and another of a signifiant assoiation with paint, laquer and glue, ontrolling for wood dust and formaldehyde. 34 The signifiant dose response observed between the risk of AD and inreasing textile dust exposure supports the existene of an assoiation found in a pooled analysis of 12 ase-ontrol studies by Lue et al. 25 This is also onsistent with the few other studies that examined the risk separately by histotype, 4 16 although their findings were not signifiant. Arseni is another well-established arinogen (group 1 IARC), 14 but only a few ase reports have been published on SNC and oupational exposure to arseni To our knowledge, this is the first epidemiologial study to report an assoiation with arseni, for whih we observed a signifiant dose response relationship. This finding requires verifiation, as risk estimates were based on few exposed ases. The absene of an effet for exposure to formaldehyde ontradits the results of the pooled analysis by Lue et al, 25 in whih a signifiant assoiation persisted after ontrolling for wood dust exposure. In a reent IARC monograph, formaldehyde was aknowledged as a definite human arinogen, but evidene of an assoiation with SNC was onsidered limited. 37 No assoiation was found between SNEC and smoking habits, in ontrast with several studies that reported relative risks ranging from 1.2 to In the present study, smoking habits ould not be preisely defined as the information on intensity and duration was inomplete for many partiipants. Nevertheless, the smoking lassifiation adopted (never, former, urrent smoker) has been shown to remove almost all onfounding by smoking when assessing the effet of oupational exposures. 39 Among the limitations of this study was the use of hospital ontrols, whih may have distorted the risk estimates if exposure to oupational hazards among ontrols was not representative of the distribution in the general population. Another soure of seletion bias ould have been the differential reporting of ases to the SNC Registry by hospitals more likely to notify the Registry of ases that had been exposed to the oupational hazards of interest, resulting in an overestimation of the risks. It is unlikely that this soure of bias seriously affeted our findings, as even among those ases reported to the Registry with probable or definite exposure to wood or leather dust, only 51% had been submitted for workers ompensation (M Slaviero, personal ommuniation, 2006). Moreover, inident ases were atively olleted by the SNC Registry and the inidene rate in this study is omparable to that reported by the network of Italian Caner Registries 40 and higher than that in the Turin Registry (the largest in Piedmont), indiating good ompleteness in data olletion. Another weakness of this study is the large number of ases exluded beause an interview on oupational history was not available (15%), though there is no obvious reason why their exlusion would be related to prior exposure to oupational hazards. Furthermore, had all 20 exluded ases been unexposed to every hazard, this would have only slightly redued the risk estimates for those exposures with low prevalene. Therefore, their exlusion is an unlikely explanation of our positive findings. Reall bias may have been a soure of differential mislassifiation of exposure, whih would have produed an overestimation of the risks. The fat that interviewers were not blind to ase-ontrol status may also have distorted risk estimates if, for example, they had been more thorough in asertaining exposure for ases than ontrols. Neither patients nor interviewers, however, had any prior knowledge of the suspeted oupational risk fators, exept perhaps for wood and leather dust. Moreover, the ourrene of SNEC was assoiated with only a few of the many hazards similarly Main messages The risk of any sino-nasal epithelial aner was signifiantly related to prior exposure to wood dust, leather dust, organi solvents, welding fumes and arseni. A dose response relationship was observed for eah of the above hazards. The hazardous exposures appear to differ by histologial type: wood dust, leather dust, organi solvents and textile dusts were risk fators for adenoarinoma, whereas exposure to welding fumes and arseni inreased the risk for squamous ell arinoma. The risk of adenoarinoma was also higher among partiipants who had been exposed to wood or leather dust at low-intensity levels. The assoiations with welding fumes, organi solvents and textile dusts merit further investigation. 454 Oup Environ Med 2009;66: doi: /oem

8 Poliy impliation The high risk for adenoarinoma with low-intensity exposure to wood dust supports the need to redue the urrent oupational exposure limit for this hazard. haraterised in the questionnaire by disernible exposure to fumes, dusts or vapours, whih did not have any effet. CONCLUSIONS Some oupational risk fators for SNEC (wood and leather dusts) were onfirmed, and dose response effets were observed for other hazards (welding fumes, organi solvents and textile dusts) that merit further investigation. The high risk for AD with low-intensity exposure to wood dust lends support for a redution in the oupational threshold value. Many of the findings by histologial type were based on few exposed ases and, therefore, the statistial adjustment for o-exposures may have been inadequate. The stability of the final models in the sensitivity analyses, however, strengthens our findings. Aknowledgements: This researh was supported by the Grant number 54-05/1999 from the Piedmont Region. We thank Drs Daniéle Lue and Frano Berrino for kindly providing us with the job-exposure matries used in the study. We are also grateful to Professor Benedetto Terraini for his invaluable omments and suggestions. Competing interests: None delared. Patient onsent: Obtained. REFERENCES 1. International Ageny for Researh on Caner. Caner in Five Continents, Vol. VIII. IARC Sientifi Publiation No Lyon: IARC, Ries LAG, Eisner MP, Kosary CL, et al, eds. SEER Caner Statistis Review, Bethesda, Maryland: National Caner Institute, sr/1975_2003/ (aessed 23 Apr 2009). 3. Aheson ED, Cowdell RH, Hadfield E, et al. Nasal aner in woodworkers in the furniture industry. BMJ 1968;2: Aheson ED, Cowdell RN, Rang E. Adenoarinoma of the nasal avity and sinuses in England and Wales. Br J Ind Med 1972;29: Hayes RB, Gerin M, Raatgever JW, et al. Wood-related oupations, wood dust exposure, and sinonasal aner. Am J Epidemiol 1986;124: Doll R, Morgan LG, Speizer FE. Caners of the lung and nasal sinuses in nikel workers. Br J Caner 1970;24: Magnus K, Andersen A, Høgetveit AC. Caner of respiratory organs among workers at a nikel refinery in Norway. Int J Caner 1982;30: Pippard EC, Aheson ED. The mortality of boot and shoe makers, with speial referene to aner. Sand J Work Environ Health 1985;11: Merler E, Baldasseroni A, Laria R, et al. On the ausal assoiation between exposure to leather dust and nasal aner: further evidene from a ase-ontrol study. Br J Ind Med 1986;43: International Ageny for Researh on Caner. IARC monographs on the evaluation of arinogeni risks to humans. Wood dust and formaldehyde. Lyon, Frane: IARC, 1995: Demers PA, Kogevinas M, Boffetta P, et al. Wood dust and sino-nasal aner: pooled reanalysis of twelve ase-ontrol studies. Am J Ind Med 1995;28: International Ageny for Researh on Caner. IARC monographs on the evaluation of arinogeni risks to humans. Chromium, nikel and welding. Lyon, Frane: IARC, 1990: International Ageny for Researh on Caner. IARC monographs on the evaluation of arinogeni risks to humans. The leather tanning and proessing industries. Lyon, Frane: IARC, 1981:25. Original artile 14. International Ageny for Researh on Caner. IARC monographs on the evaluation of arinogeni risks to humans, Supplement 7. Overall evaluations of arinogeniity: an updating of IARC monographs Volumes 1 to 42. Lyon, Frane: IARC, Lue D, Leler A, Moret JF, et al. Oupational risk fators for sinonasal aner: a ase-ontrol study in Frane. Am J Ind Med 1992;21: Brinton LA, Blot WJ, Fraumeni JF Jr. Nasal aner in the textile and lothing industries. Br J Ind Med 1985;42: Lue D, Gérin M, Moret JF, et al. Sinonasal aner and oupational exposure to textile dust. Am J Ind Med 1997;32: Brinton LA, Blot WJ, Beker JA, et al. A ase-ontrol study of aners of the nasal avity and paranasal sinuses. Am J Epidemiol 1984;119: Leler A, Martinez Cortes M, Gérin M, et al. Sinonasal aner and wood dust exposure: results from a ase-ontrol study. Am J Epidemiol 1994;140: Hernberg S, Westerholm P, Shultz-Larsen K, et al. Nasal and sinonasal aner: onnetion with oupational exposure in Denmark, Finland and Sweden. Sand J Work Environ Health 1983;9: Olsen JH. Oupational risks of sinonasal aner in Denmark. Br J Ind Med 1988;45: Vaughan TL. Nasal aner in wood-related industries. JOM 1989;31: Lue D, Gérin M, Berrino F, et al. Soures of disrepanies between a job exposure matrix and a ase by ase expert assessment for oupational exposure to formaldehyde and wood-dust. Int J Epidemiol 1993;22(Suppl 2):S Berrino F, Rihiardi L, Boffetta P, et al. Oupation and larynx and hypopharynx aner: a job-exposure matrix approah in an international ase-ontrol study in Frane, Italy, Spain and Switzerland. Caner Causes Control 2003;14: Lue D, Leler A, Bégin D, et al. Sinonasal aner and oupational exposures: a pooled analysis of 12 ase-ontrol studies. Caner Causes Control 2002;13: Choi BC, Howe GR. Methodologial issues in ase-ontrol studies IV: Validity and effiieny of various analysis strategies for ontinuous variables using the unonditional logisti regression model. Int J Epidemiol 1984;13: Hosmer DW, Lemeshow S. Applied logisti regression. 2nd edn. New York: John Wiley & Sons, Kauppinen TP, Mutanen PO, Seitsamo JT. Magnitude of mislassifiation bias when using a job-exposure matrix. Sand J Work Environ Health 1992;18: NIOSH. Reommendations for oupational safety and health: Compendium of poliy douments and statements. Cininnati, Ohio: U.S. Department of Health and Human Servies, Publi Health Servie, Centers for Disease Control, National Institute for Oupational Safety and Health, DHHS (NIOSH) Publiation No , Comba P, Battista G, Belli S, et al. A ase-ontrol study of aner of the nose and paranasal sinuses and oupational exposures. Am J Ind Med 1992;22: Shimizu H, Hozawa J, Saito H, et al. Chroni sinusitis and woodworking as risk fators for aner of the maxillary sinus in Northeast Japan. Laryngosope 1989;99: Teshke K, Morgan MS, Chekoway H, et al. Surveillane of nasal and bladder aner to loate soures of exposure to oupational arinogens. Oup Environ Med 1997;54: Ambroise D, Wild P, Moulin JJ. Update of a meta-analysis on lung aner and welding. Sand J Work Environ Health 2006;32: Olsen JH, Jensen SP, Hink M, et al. Oupational formaldehyde exposure and inreased nasal aner risk in man. Int J Caner 1984;34: Roth F. Uber den Bronhialkrebs arsengeshadigter Winzer. Virhows Arh Pathol Anat 1958;331: [Artile in German.] 36. Battista G, Bartoli D, Iaia TE, et al. Art glassware and sinonasal aner: report of three ases. Am J Ind Med 1996;30: International Ageny for Researh on Caner. IARC monographs on the evaluation of arinogeni risks to humans. Vol. 88: Formaldehyde, 2-butoxyethanol and 1-tert-butoxypropan-2-ol. Lyon, Frane: IARC, Littman AJ, Vaughan TL. Caners of the nasal avity and paranasal sinuses. In: Shottenfeld D, Fraumeni JF Jr, eds. Caner epidemiology and prevention. New York: Oxford University Press, 2006: Rihiardi L, Forastiere F, Boffetta P, et al. Effet of different approahes to treatment of smoking as a potential onfounder in a ase-ontrol study on oupational exposures. Oup Environ Med 2005;62: AIRT. I dati di inidenza e mortalità dei Registri Tumori generali, Dati integrativi alla monografia, [In Italian.] inidenza /doumenti/tassi%20grezzi.pdf (aessed 23 Apr 2009). Oup Environ Med 2009;66: doi: /oem